Quantitative Inversion Of Biochemical Compositions By Hyperspectral Remote Sensing

Vegetation is one of the most important components of the ecosystem with 70 percent of the land surface coverage. Components such as chlorophyll, water, protein, lignin, cellulose and elements including carbon, nitrogen, hydrogen in the vegetation are defined as the vegetation biochemistry composition. The content and structure of these compositions can be directly or indirectly influenced on the ecological and physiological processing which is important to human being activities. As hot point and frontier in remote sensing, hyperspectral remote sensing technique not only has the advantages of traditional remote sensing that can be timely and nondestructively used to detect large vegetation area, but also has special advantages with the very high spectral resolution. More delicate spectral difference can help us to accurately retrieve vegetation biochemistry compositions and to monitor. Vegetation biochemistry extraction based on the remote sensing So composition estimation based on hyperspectral remote sensing technique is urgent not only for study in ecology, agronomy and global change but also for application in yield predict and precision farming.This thesis focuses on vegetation components retrieval with remote sensing especially hyperspectral remote sensing, and puts great emphasis on the study of extracting canopy biochemistry from the satellite-based hyperspectral imagery. The first chapter mainly introduced the principle, method and experiment basis of hyperspectral remote sensing in vegetation biochemistry, and then the emphasis of the article. In the second chapter, we primarily introduced the data obtainment, particularly discussed the satellite-land simultaneous experiment of XiShuangBanNa in Yunnan province. The experiment is included the design, sample collection, spectra measurement and imagery order. The third chapter is a foundation research for the canopy biochemistry retrieval. We mainly consider the leaf-level extraction using the hyperspectral response characteristics. The fourth part studied the hyperspectral imagery processing including the geometrical registration and atmospheric correction. The fifth part is the most important in this thesis. The relationship between the Hyperion spectra and the biochemical components is established and further obtain its distribution maps.The last chapter summarized the whole thesis and listed the achievement of this study, as same as, pointed out the research direction in the future. Main development and conclusion as follows:1. Satellite-land simultaneous experiment of XiShuangBanNa in Yunnan province collected the biochemical, measured spectra and hyperspectral remote sensing imagery data. We summarized the emphases and proposed some reasonable suggest for the related experiment. It indicated the experimental feature.2. Using spectral position (wavelength) variable analysis technique, two parameters, namely, first derivative extremum and area normalized first derivative extremum are brought forward and employed to retrieve leaf total nitrogen, cellulose, lignin, starch and water content with good results. Research indicates that the parameters can effectively remove the background influence.3. Leaf water content is retrieved and tested using the LOPEX dataset. The research results indicate that model inversion precision can be improved based on the leaf type and water expression. The spectra indices SR and Ratio975 is the best for FMC and EWT extraction, respectively.4. Simulation results by ACRM model exhibit that the short-wave infrared not the near-infrared wavelength is sensitivity to water content, namely the 1600nm and 820nm. So the two bands combined can strengthen the vegetation water content information. A soil adjusted water index (SAWI) based on the two bands is proposed to compute the canopy water content. SAWI simulation results under different LAI and Cw, different LAI and N combination showed that the index is a possibly applicable in water calculation.5. Hyperspectral imagery processing is studied including geometrical registration and atmospheric correction. Atmospheric correction of Hyperion image is managed both the dark/bright empirical method and 6S physical model. The key of the former is the selection of the calibration objects which need large enough, near Lambert and short of vegetation cover. 6S model atmospheric correction based on MOD IS data can overcome the difficult obtainments of water vapor, ozone concentration and aerosol optical depth. The two methods achieve the satisfactory correction accuracy.Through the partial least squares method, canopy chlorophyll, water, total carbon, total nitrogen, hydrogen and carbon-nitrogen ratio are retrieved and mapped using the processed Hyperion image and measured biochemistry data. It indicated that 1) the carbon-nitrogen ratio in this area is around 30;2) the reflectance in the wavelength of 2052nm and 2173nm is sensitive to canopy total nitrogen, carbon-nitrogen ratio and carbon content. From the practical point of view, using the sensitive bands can produce the instruments to detect the canopy nitrogen and carbon-nitrogen ratio timely and nondestructively. This part established a good foundation for vegetation hyperspectral remote sensing and carbon or nitrogen cycle research in our southwest tropical and sub-tropical humid climate.